@ propbeanie
Thank you for the suggestions mate. Yes, definitely there is an overlapping that needs to be observed between one tile and the next ones. Reason for that is that each pixel on those maps doesn't represent a quadrant in a grid, but a verted in a continuos mesh, and at least one row/column of vertices must be in common for two neighbouring terrain tiles for them to be treated as a continuous surface. If their heigh/grey values are not identical, an abrupt height change or a step is created, that from the game perspective are like a trench ir a wall.
That said, I think that SH5 tiles have an overlapping of just one pixel. Why I think that?
1 - by comparing stock tiles with real world terrain height maps, if a 6-pixel overlapping was to be applied, land masses on those tiles should look a bit smaller than they do, whereas they actually show the same land proportions as seen in the big (3601x3601 pixels)
srtm maps
2 - as I said before, 601 pixels - 1 pixel of overlapping makes a resolution of exactly one height point every six arcseconds (= 200 m in SH5). A bigger overlapping would involve a decimal resolution which seems not likely to me.
3 - Before starting with my editing I had exporter two adjoining maps and overlapped them using the "divide" blending mode. The division of two identical pixels is equal to pure white (any number divided by itself makes 1 which, in image processing, is equal to white). The only way I could get all white pixels was by setting an overlap of only one pixel
I will do more test, but I think that my working method is correct for SH5. Yes, editing those maps requires a general plan. The error is around the corner, but as far as you keep track of all the changes made and you save every change in separate layers (that's easy in photoshop), keeping backup copies of the edited files, You don't necessarily need to edit everything at once. The four extra bits added by Ref's exporter are a bit annoying, but if you have saved copies of your working files in psd format, that's not a big problem either
